12 Apr Two NVIDIA GTC Automotive Announcements That Deserve More Attention
As I wrote in my lead-up to NVIDIA’s 2019 GPU Technology Conference (GTC), the annual event is an industry “who’s who” in AI and Deep Learning and is not one to miss. As usual, it did not disappoint—NVIDIA showcased a multitude of impactful applications of its technology, from data center and professional visualization to autonomous vehicles and robotics. There were a few announcements from the event, however, that I felt didn’t get their due attention. I wanted to highlight those today.
Expanded relationship with Toyota
The first takeaway that I think flew under the radar is that Toyota is now “all-in” on NVIDIA. Together, NVIDIA and the Toyota Research Institute-Advanced Development (TRI-AD) team announced a new collaboration on the development, training, and validation of autonomous vehicles. The companies both say the collaboration is broad, encompassing the development of AI computing infrastructure utilizing NVIDIA GPUs, advancements in simulation with the NVIDIA DRIVE Constellation Platform (read more about DRIVE Constellation here and here), and in-car AV computers based on DRIVE AGX Pegasus and Xavier.
As a firm, we’ve been keeping a close eye on NVIDIA’s aggressive efforts in the autonomous driving market. AI is NVIDIA’s big differentiator and happens to be essential for enabling self-driving cars to take in information from their environments and make split-second decisions. Simulation is also huge in AV, preparing vehicles for any number of real-world scenarios and ensuring safety before they hit the (actual) roads. With its two side-by-side servers, DRIVE Constellation effectively simulates the equivalent of billions of miles of driving. While this technology is impressive in its rite, it takes partnerships with automakers like Toyota to fully reap the benefits and bring AVs to market. This end-to-end partnership between the world’s leading automaker (number #2 in sales behind VW in 2018) and the tech industry’s leader in AI and HPC should bear impressive fruit.
DRIVE Constellation now available
The second (and related) piece of news from GTC 2019 was that DRIVE Constellation, originally announced at GTC last year, is now available. In brief, DRIVE Constellation is a datacenter solution comprised of two servers: DRIVE Constellation Simulator, which generates sensor output from the virtual car, and DRIVE Constellation Vehicle, which processes the data (via the DRIVE AGX Pegasus AI Car Computer) and feeds it back into the simulator. This is machines (DGX) teaching other machines on how to drive other machines (Constellation) grading the exam to ensure the system gets a perfect score before it graduates.
Imagine the ability to simulate virtually any adverse weather scenario—rain, ice, snow—on a variety of road surfaces, any time of day or night. Another important aspect to DRIVE Constellation is that it is an “open” platform, open in that it is capable of incorporating datasets from the simulation ecosystem at large (for example, a simulation company called Cognata announced at GTC that its scenario and traffic model could be supported on DRIVE). All of this to say that DRIVE Constellation is the future of testing and I think will be huge from an industry impact point of view. AVs are going to be one of the most impactful technologies of the modern age, but to do so they must first and foremost be safe. NVIDIA has one of the most powerful simulators in the market, and it’s open. I think it’s going to play a huge role in testing and validating the next-generation of AVs.
There was a lot of news and analysis that came out of GTC 2019, and both Toyota and Constellation didn’t get the attention they deserved. The takeaway from both of these announcements is that NVIDIA’s efforts in the AV market are in a constant state of growth along the journey. The AV market is new, has no rulebook and companies who participate need to be able to move to where the market is, which seems to change annually. I think DRIVE Constellation is going to be the go-to approach for training as it can simulate billions of miles–and more importantly potentially hazardous scenarios– driven in many different conditions, and Toyota is a huge first customer win for the platform. I’ll continue to watch with interest.
Note: Moor Insights & Strategy writers and editors may have contributed to this article.